Concentration And Velocity Fields Throughout The Flow Field Of Swirling Flows In Gas Turbine Mixers

Transcription

1 University of Central Florida Electronic Theses and Dissertations Doctoral Dissertation (Open Access) Concentration And Velocity Fields Throughout The Flow Field Of Swirling Flows In Gas Turbine Mixers 2004 Louis James Turek University of Central Florida Find similar works at: University of Central Florida Libraries Part of the Engineering Commons STARS Citation Turek, Louis James, "Concentration And Velocity Fields Throughout The Flow Field Of Swirling Flows In Gas Turbine Mixers" (2004). Electronic Theses and Dissertations This Doctoral Dissertation (Open Access) is brought to you for free and open access by STARS. It has been accepted for inclusion in Electronic Theses and Dissertations by an authorized administrator of STARS. For more information, please contact

2 CONCENTRATION AND VELOCITY FIELDS THROUGHOUT THE FLOW FIELD OF SWIRLING FLOWS IN GAS TURBINE MIXERS by LOUIS JAMES TUREK B.S. University of Central Florida, 2000 M.S. University of Central Florida, 2001 A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Mechanical, Materials, and Aerospace Engineering in the College of Engineering and Computer Science at the University of Central Florida Orlando, Florida Summer Term 2004 Major Professor: Ruey-Hung Chen

3 ABSTRACT Air velocity and fuel concentration data have been collected throughout the flow fields of two gas turbine mixers in an effort to better understand the mixing of fuel and air in gas turbine mixers. The two gas turbine mixers consisted of an annular flow profile and incorporated swirl vanes to produce a swirling flow to promote fuel/air mixing. The fuel was injected into the bulk flow from the pressure side of the swirl vanes. The first mixer had a swirl angle of 45 o, while the second had a swirl angle of 55 o. In order to examine the effect of the swirl angle on the mixing of fuel and air as the flow progressed through gas turbine mixers, axial and tangential air velocity data was taken using a laser Doppler velocimeter (LDV). Also, fuel concentration data was taken separately using a hydrocarbon concentration probe with methane diluted with air as the fuel. The data were taken at varying axial and varying angular locations in an effort to capture the spatial development of the fuel and velocity profiles. The spectra of the data were analyzed as well in an effort to understand the turbulence of the flow. It was found that the 55 o swirler exhibited smaller variations in both velocity and fuel concentration values and that the fuel reached a uniform concentration at axial locations further upstream in the 55 o degree mixer than in the 45 o mixer. The RMS values of the velocity, which were influenced by the swirl vanes, were higher in the 55 o mixer and likely contributed to the better mixing performance of the 55 o mixer. The fuel concentration spectrum data showed that the spectra of the two mixers were similar, and that the fluctuations in fuel concentration due to flow emanating from the swirl vanes were seen throughout the length of the two mixers. ii

4 ACKNOWLEDGEMENTS The author would like to gratefully acknowledge the technical assistance of Dr. Marcos Chaos, Mr. Anupam Kothawala, and Mr. Chris Douglass in the setup of equipment and hardware in the work discussed herein. The advice and support of the committee members, particularly the committee chair Dr. Ruey-Hung Chen, is also greatly appreciated. The author is also thankful for the donation by Siemens- Westinghouse Power Corporation of the hardware examined in this work. Also, appreciation for the love and support of my parents, Anna and Lou Turek, is beyond what can be expressed in mere words. iii

5 TABLE OF CONTENTS LIST OF FIGURES... v LIST OF TABLES... xvi LIST OF SYMBOLS... xvii INTRODUCTION Background Previous Work EXPERIMENTAL METHODS AND SETUP Laser Doppler Velocimeter Hydrocarbon Concentration Probe RESULTS Velocity Data Axial Velocity vs. θ Axial RMS Velocity vs. θ Tangential Velocity vs. θ Tangential RMS Velocity vs. θ Axial Velocity vs Axial RMS Velocity vs Tangential Velocity vs Tangential RMS Velocity vs Fuel Concentration Data Fuel Concentration vs. θ RMS Fuel Concentration vs. θ Fuel Concentration vs RMS Fuel Concentration vs Unmixedness Unmixedness at θ= 0 degrees Sector Averaged Unmixedness Spectrum Analysis Axial Velocity Spectrum Tangential Velocity Spectrum Fuel Concentration Spectrum CONCLUSIONS LIST OF REFERENCES APPENDIX iv

6 LIST OF FIGURES Figure 1. Diagram of swirl vane geometric parameters Figure 2. Test rig with swirler mounted into acrylic tube Figure 3. Test hardware with lasers passing through antireflective glass Figure 4. Diagram of LDV and test rig Figure 5. Module 3 mounted in test rig Figure 6. Module 4, with optically clear aft shroud, mounted in test rig Figure 7. Drawing of laser crossing with dimensions diagramed Figure 8. Typical Doppler burst during velocity data collection Figure 9. Histogram showing data points clustered around the mean velocity value (33.39 m/s) Figure 10. Diagram of data acquisition equipment connections Figure 11. Diagram of swirler nozzle with titanium dioxide supply system Figure 12. Diagram of concentration probe hardware Figure 13. Diagram of fuel flow system with test hardware Figure 14. FFT plots at with frequency of (a) 50 Hz and (b) 550 Hz Figure 15. FFT plot a frequency too high (600 Hz) to be resolved by concentration probe Figure 16. Data collection locations for Module 3 (a) and Module 4 (b) Figure 17. Mean axial velocity at = 0.87 for Module Figure 18. Mean tangential velocity at = 0.87 for Module Figure 19. Mean axial velocity at = for Module Figure 20. Mean tangential velocity at = for Module Figure 21. Radial profiles of normalized mean axial velocity vs. θ, = 0.89 for Module Figure 22. Radial profiles of normalized mean axial velocity vs. θ, = 0.87 for Module Figure 23. Radial profiles of normalized mean axial velocity vs. θ, = 1.11 for Module Figure 24. Radial profiles of normalized mean axial velocity vs. θ, = 1.06 for Module Figure 25. Radial profiles of normalized mean axial velocity vs. θ, = 1.32 for Module Figure 26. Radial profiles of normalized mean axial velocity vs. θ, = 1.25 for Module Figure 27. Radial profiles of normalized mean axial velocity vs. θ, = 1.74 for Module Figure 28. Radial profiles of normalized mean axial velocity vs. θ, = 1.63 for Module Figure 29. Radial profiles of normalized mean axial velocity vs. θ, = 2.17 for Module Figure 30. Radial profiles of normalized mean axial velocity vs. θ, = 2.02 for Module v

7 Figure 31. Radial profiles of normalized mean axial velocity vs. θ, = for Module Figure 32. Radial profiles of normalized mean axial velocity vs. θ, = 2.21 for Module Figure 33. Radial profiles of normalized mean axial velocity vs. θ, = 3.02 for Module Figure 34. Radial profiles of normalized mean axial velocity vs. θ, = for Module Figure 35. Radial profiles of sector averaged normalized mean axial velocity for Module Figure 36. Radial profiles of sector averaged normalized mean axial velocity for Module Figure 37. Radial profiles of normalized axial RMS velocity vs. θ, = 0.89 for Module Figure 38. Radial profiles of normalized axial RMS velocity vs. θ, = 0.87 for Module Figure 39. Radial profiles of normalized axial RMS velocity vs. θ, = 1.11 for Module Figure 40. Radial profiles of normalized axial RMS velocity vs. θ, = 1.06 for Module Figure 41. Radial profiles of normalized axial RMS velocity vs. θ, = 1.32 for Module Figure 42. Radial profiles of normalized axial RMS velocity vs. θ, = 1.25 for Module Figure 43. Radial profiles of normalized axial RMS velocity vs. θ, = 1.74 for Module Figure 44. Radial profiles of normalized axial RMS velocity vs. θ, = 1.63 for Module Figure 45. Radial profiles of normalized axial RMS velocity vs. θ, = 2.17 for Module Figure 46. Radial profiles of normalized axial RMS velocity vs. θ, = 2.02 for Module Figure 47. Radial profiles of normalized axial RMS velocity vs. θ, = for Module Figure 48. Radial profiles of normalized axial RMS velocity vs. θ, = 2.21 for Module Figure 49. Radial profiles of normalized axial RMS velocity vs. θ, = 3.02 for Module Figure 50. Radial profiles of normalized axial RMS velocity vs. θ, = for Module Figure 51. Radial profiles of sector averaged normalized axial RMS velocity for Module Figure 52. Radial profiles of sector averaged normalized axial RMS velocity for Module vi

8 Figure 53. Radial profiles of normalized mean tangential velocity vs. θ, = 0.89 for Module Figure 54. Radial profiles of normalized mean tangential velocity vs. θ, = 0.87 for Module Figure 55. Radial profiles of normalized mean tangential velocity vs. θ, = 1.11 for Module Figure 56. Radial profiles of normalized mean tangential velocity vs. θ, = 1.06 for Module Figure 57. Radial profiles of normalized mean tangential velocity vs. θ, = 1.32 for Module Figure 58. Radial profiles of normalized mean tangential velocity vs. θ, = 1.25 for Module Figure 59. Radial profiles of normalized mean tangential velocity vs. θ, = 1.74 for Module Figure 60. Radial profiles of normalized mean tangential velocity vs. θ, = 1.63 for Module Figure 61. Radial profiles of normalized mean tangential velocity vs. θ, = 2.17 for Module Figure 62. Radial profiles of normalized mean tangential velocity vs. θ, = 2.02 for Module Figure 63. Radial profiles of normalized mean tangential velocity vs. θ, = for Module Figure 64. Radial profiles of normalized mean tangential velocity vs. θ, = 2.21 for Module Figure 65. Radial profiles of normalized mean tangential velocity vs. θ, = 3.02 for Module Figure 66. Radial profiles of normalized mean tangential velocity vs. θ, = for Module Figure 67. Radial profiles of sector averaged normalized mean tangential velocity for Module Figure 68. Radial profiles of sector averaged normalized mean tangential velocity for Module Figure 69. Radial profiles of normalized tangential RMS velocity vs. θ, = 0.89 for Module Figure 70. Radial profiles of normalized tangential RMS velocity vs. θ, = 0.87 for Module Figure 71. Radial profiles of normalized tangential RMS velocity vs. θ, = 1.11 for Module Figure 72. Radial profiles of normalized tangential RMS velocity vs. θ, = 1.06 for Module Figure 73. Radial profiles of normalized tangential RMS velocity vs. θ, = 1.32 for Module Figure 74. Radial profiles of normalized tangential RMS velocity vs. θ, = 1.25 for Module vii

9 Figure 75. Radial profiles of normalized tangential RMS velocity vs. θ, = 1.74 for Module Figure 76. Radial profiles of normalized tangential RMS velocity vs. θ, = 1.63 for Module Figure 77. Radial profiles of normalized tangential RMS velocity vs. θ, = 2.17 for Module Figure 78. Radial profiles of normalized tangential RMS velocity vs. θ, = 2.02 for Module Figure 79. Radial profiles of normalized tangential RMS velocity vs. θ, = for Module Figure 80. Radial profiles of normalized tangential RMS velocity vs. θ, = 2.21 for Module Figure 81. Radial profiles of normalized tangential RMS velocity vs. θ, = 3.02 for Module Figure 82. Radial profiles of normalized tangential RMS velocity vs. θ, = for Module Figure 83. Radial profiles of sector averaged normalized tangential RMS velocity for Module Figure 84. Radial profiles of sector averaged normalized tangential RMS velocity for Module Figure 85. Radial profiles of normalized mean axial velocity vs., θ= 0 deg for Module Figure 86. Radial profiles of normalized mean axial velocity vs., θ= 0 deg for Module Figure 87. Radial profiles of normalized mean axial velocity vs., θ= 6 deg for Module Figure 88. Radial profiles of normalized mean axial velocity vs., θ= 6 deg for Module Figure 89. Radial profiles of normalized mean axial velocity vs., θ= 12 deg for Module Figure 90. Radial profiles of normalized mean axial velocity vs., θ= 12 deg for Module Figure 91. Radial profiles of normalized mean axial velocity vs., θ= 18 deg for Module Figure 92. Radial profiles of normalized mean axial velocity vs., θ= 18 deg for Module Figure 93. Radial profiles of normalized mean axial velocity vs., θ= 24 deg for Module Figure 94. Radial profiles of normalized mean axial velocity vs., θ= 24 deg for Module Figure 95. Radial profiles of normalized mean axial velocity vs., θ= 30 deg for Module Figure 96. Radial profiles of normalized mean axial velocity vs., θ= 30 deg for Module viii

10 Figure 97. Radial profiles of normalized axial RMS velocity vs., θ= 0 deg for Module Figure 98. Radial profiles of normalized axial RMS velocity vs., θ= 0 deg for Module Figure 99. Radial profiles of normalized axial RMS velocity vs., θ= 6 deg for Module Figure 100. Radial profiles of normalized axial RMS velocity vs., θ= 6 deg for Module Figure 101. Radial profiles of normalized axial RMS velocity vs., θ= 12 deg for Module Figure 102. Radial profiles of normalized axial RMS velocity vs., θ= 12 deg for Module Figure 103. Radial profiles of normalized axial RMS velocity vs., θ= 18 deg for Module Figure 104. Radial profiles of normalized axial RMS velocity vs., θ= 18 deg for Module Figure 105. Radial profiles of normalized axial RMS velocity vs., θ= 24 deg for Module Figure 106. Radial profiles of normalized axial RMS velocity vs., θ= 24 deg for Module Figure 107. Radial profiles of normalized axial RMS velocity vs., θ= 30 deg for Module Figure 108. Radial profiles of normalized axial RMS velocity vs., θ= 30 deg for Module Figure 109. Radial profiles of normalized mean tangential velocity vs., θ= 0 deg for Module Figure 110. Radial profiles of normalized mean tangential velocity vs., θ= 0 deg for Module Figure 111. Radial profiles of normalized mean tangential velocity vs., θ= 6 deg for Module Figure 112. Radial profiles of normalized mean tangential velocity vs., θ= 6 deg for Module Figure 113. Radial profiles of normalized mean tangential velocity vs., θ= 12 deg for Module Figure 114. Radial profiles of normalized mean tangential velocity vs., θ= 12 deg for Module Figure 115. Radial profiles of normalized mean tangential velocity vs., θ= 18 deg for Module Figure 116. Radial profiles of normalized mean tangential velocity vs., θ= 18 deg for Module Figure 117. Radial profiles of normalized mean tangential velocity vs., θ= 24 deg for Module Figure 118. Radial profiles of normalized mean tangential velocity vs., θ= 24 deg for Module ix

11 Figure 119. Radial profiles of normalized mean tangential velocity vs., θ= 30 deg for Module Figure 120. Radial profiles of normalized mean tangential velocity vs., θ= 30 deg for Module Figure 121. Radial profiles of normalized tangential RMS velocity vs., θ= 0 deg for Module Figure 122. Radial profiles of normalized tangential RMS velocity vs., θ= 0 deg for Module Figure 123. Radial profiles of normalized tangential RMS velocity vs., θ= 6 deg for Module Figure 124. Radial profiles of normalized tangential RMS velocity vs., θ= 6 deg for Module Figure 125. Radial profiles of normalized tangential RMS velocity vs., θ= 12 deg for Module Figure 126. Radial profiles of normalized tangential RMS velocity vs., θ= 12 deg for Module Figure 127. Radial profiles of normalized tangential RMS velocity vs., θ= 18 deg for Module Figure 128. Radial profiles of normalized tangential RMS velocity vs., θ= 18 deg for Module Figure 129. Radial profiles of normalized tangential RMS velocity vs., θ= 24 deg for Module Figure 130. Radial profiles of normalized tangential RMS velocity vs., θ= 24 deg for Module Figure 131. Radial profiles of normalized tangential RMS velocity vs., θ= 30 deg for Module Figure 132. Radial profiles of normalized tangential RMS velocity vs., θ= 30 deg for Module Figure 133. Radial profiles of normalized mean CH 4 concentration vs. θ, = 0.89 for Module Figure 134. Radial profiles of normalized mean CH 4 concentration vs. θ, = 0.87 for Module Figure 135. Radial profiles of normalized mean CH 4 concentration vs. θ, = 1.11 for Module Figure 136. Radial profiles of normalized mean CH 4 concentration vs. θ, = 1.06 for Module Figure 137. Radial profiles of normalized mean CH 4 concentration vs. θ, = 1.32 for Module Figure 138. Radial profiles of normalized mean CH 4 concentration vs. θ, = 1.25 for Module Figure 139. Radial profiles of normalized mean CH 4 concentration vs. θ, = 1.74 for Module Figure 140. Radial profiles of normalized mean CH 4 concentration vs. θ, = 1.63 for Module x

12 Figure 141. Radial profiles of normalized mean CH 4 concentration vs. θ, = 2.17 for Module Figure 142. Radial profiles of normalized mean CH 4 concentration vs. θ, = 2.02 for Module Figure 143. Radial profiles of normalized mean CH 4 concentration vs. θ, = for Module Figure 144. Radial profiles of normalized mean CH 4 concentration vs. θ, = 2.21 for Module Figure 145. Radial profiles of normalized mean CH 4 concentration vs. θ, = 3.02 for Module Figure 146. Radial profiles of normalized mean CH 4 concentration vs. θ, = for Module Figure 147. Radial profiles of sector averaged normalized mean CH 4 concentration for Module Figure 148. Radial profiles of sector averaged normalized mean CH 4 concentration for Module Figure 149. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 0.89 for Module Figure 150. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 0.87 for Module Figure 151. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 1.11 for Module Figure 152. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 1.06 for Module Figure 153. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 1.32 for Module Figure 154. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 1.25 for Module Figure 155. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 1.74 for Module Figure 156. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 1.63 for Module Figure 157. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 2.17 for Module Figure 158. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 2.02 for Module Figure 159. Radial profiles of normalized RMS CH 4 concentration vs. θ, = for Module Figure 160. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 2.21 for Module Figure 161. Radial profiles of normalized RMS CH 4 concentration vs. θ, = 3.02 for Module Figure 162. Radial profiles of normalized RMS CH 4 concentration vs. θ, = for Module xi

13 Figure 163. Radial profiles of sector averaged normalized RMS CH 4 concentration for Module Figure 164. Radial profiles of sector averaged normalized RMS CH 4 concentration for Module Figure 165. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 166. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 167. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 168. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 169. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 170. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 171. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 172. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 173. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 174. Radial profiles of normalized mean CH 4 concentration vs., θ= for Module Figure 175. Radial profiles of normalized mean CH 4 concentration vs., θ= 3 for Module Figure 176. Radial profiles of normalized mean CH 4 concentration vs., θ= 3 for Module Figure 177. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 178. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 179. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 180. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 181. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 182. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 183. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 184. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module xii

14 Figure 185. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 186. Radial profiles of normalized RMS CH 4 concentration vs., θ= for Module Figure 187. Radial profiles of normalized RMS CH 4 concentration vs., θ= 3 for Module Figure 188. Radial profiles of normalized RMS CH 4 concentration vs., θ= 3 for Module Figure 189. Unmixedness at θ=, = 0.89 for Module 3 and = 0.87 for Module Figure 190. Unmixedness at θ=, = 1.74 for Module 3 and = 1.63 for Module Figure 191. Unmixedness at θ=, = for Module 3 and = for Module Figure 192. Sector averaged unmixedness at = 0.89 for Module 3 and = 0.87 for Module Figure 193. Sector averaged unmixedness at = 1.74 for Module 3 and = 1.63 for Module Figure 194. Sector averaged unmixedness at = for Module 3 and = for Module Figure 195. Axial velocity spectrum vs. for Module 3, θ= Figure 196. Axial velocity spectrum vs. for Module 4, θ= Figure 197. Axial velocity spectrum vs. for Module 3, θ= Figure 198. Axial velocity spectrum vs. for Module 4, θ= Figure 199. Axial velocity spectrum vs. for Module 3, θ= Figure 200. Axial velocity spectrum vs. for Module 4, θ= Figure 201. Axial velocity spectrum vs. for Module 3, θ= Figure 202. Axial velocity spectrum vs. for Module 4, θ= Figure 203. Axial velocity spectrum vs. for Module 3, θ= Figure 204. Axial velocity spectrum vs. for Module 4, θ= Figure 205. Axial velocity spectrum vs. for Module 3, θ= Figure 206. Axial velocity spectrum vs. for Module 4, θ= Figure 207. Tangential velocity spectrum vs. for Module 3, θ= Figure 208. Tangential velocity spectrum vs. for Module 4, θ= Figure 209. Tangential velocity spectrum vs. for Module 3, θ= Figure 210. Tangential velocity spectrum vs. for Module 4, θ= Figure 211. Tangential velocity spectrum vs. for Module 3, θ= Figure 212. Tangential velocity spectrum vs. for Module 4, θ= Figure 213. Tangential velocity spectrum vs. for Module 3, θ= Figure 214. Tangential velocity spectrum vs. for Module 4, θ= Figure 215. Tangential velocity spectrum vs. for Module 3, θ= Figure 216. Tangential velocity spectrum vs. for Module 4, θ= Figure 217. Tangential velocity spectrum vs. for Module 3, θ= Figure 218. Tangential velocity spectrum vs. for Module 4, θ= xiii

15 Figure 219. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 220. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 221. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 222. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 223. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 224. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 225. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 226. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 227. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 228. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 229. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 230. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 231. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 232. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 233. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 234. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 235. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 236. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 237. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 238. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 239. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 240. CH 4 concentration spectrum for Module 4 vs. at θ=, = xiv

16 Figure 241. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 242. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 243. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 244. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 245. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 246. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 247. CH 4 concentration spectrum for Module 3 vs. at θ=, = Figure 248. CH 4 concentration spectrum for Module 4 vs. at θ=, = Figure 249. CH 4 concentration spectrum for Module 3 vs. at θ= 3, = Figure 250. CH 4 concentration spectrum for Module 4 vs. at θ= 3, = Figure 251. CH 4 concentration spectrum for Module 3 vs. at θ= 3, = Figure 252. CH 4 concentration spectrum for Module 4 vs. at θ= 3, = Figure 253. CH 4 concentration spectrum for Module 3 vs. at θ= 3, = Figure 254. CH 4 concentration spectrum for Module 4 vs. at θ= 3, = Figure 255. Mean axial velocity at = 0.87 for Module 4 with ± 4.9% error bar Figure 256. Mean tangential velocity at = 0.87 for Module 4 with ± 4.9% error bar Figure 257. Mean axial velocity at = for Module 4 with ± 4.9% error bar Figure 258. Mean tangential velocity at = for Module 4 with ± 4.9% error bar xv

17 LIST OF TABLES Table 1. Geometric parameters of swirler nozzles Table 2. Geometric parameters of swirl vanes Table 3. LDV parameters (see Figure 7) Table 4. Axial locations for data collection xvi

18 LIST OF SYMBOLS a mean particle diameter, µm C methane concentration measured by concentration probe C f methane concentration in methane/air mixture at fuel injection holes c v length of swirl vane along chord line, mm d f fringe spacing, µm d m probe volume diameter, mm l m probe volume length, mm D a diameter of receiving optics lens D e2 diameter of laser beams, mm G particle scattering parameter f focal length of LDV optics, mm f k frequency of Kolmogorov eddy fluctuation, Hz f m frequency of integral scale eddy fluctuation, Hz h v height of swirl vane from leading edge to trailing edge, mm L length of absorption path in concentration probe, mm l camber length of swirl vane along camber line, mm P L power in each laser beam of LDV, watts P o reference pressure for concentration measurements, atm P total total pressure of sample for concentration measurements, atm R universal gas constant, (atm cm 3 )/(mol K) r a focal length of receiving optics lens, mm SNR signal to noise ratio, dimensionless T temperature, K t max maximum thickness of swirl vane, mm Un Uncertainty of methane concentration measurements, percent V velocity, m/s V bulk bulk velocity, m/s V i visibility of particles χ molar concentration f bandwidth of signal, MHz ε decadic molar extinction coefficient φ equivalence ratio, dimensionless κ half angle between laser beams, degrees η k size of Kolmogorov eddies, mm η q quantum efficiency of photodetector, dimensionless λ wavelength of laser light, nm θ angle of module orientation, degrees infrared transmittance of the sample τ IR xvii

19 INTRODUCTION In the combustion processes of gas turbines, extensive research has been done in an effort to minimize certain chemical species in the exhaust gases. These chemical species are sources of pollution in the atmosphere, and thus are an undesirable result of the combustion process. Two chemical species that have been the focus of a great deal of effort are nitric oxide (NO) and nitrogen dioxide (NO 2 ), collectively referred to as NO x. The importance of minimizing NO x production is discussed by Seinfeld [1], and stems largely from the effect that NO x has on ozone molecules in the atmosphere. Seinfeld [1] discussed the destruction of ozone by NO x, which is the most important destruction process for ozone molecules, and is given by Equations (1) and (2) below: O+NO 2 NO+O 2 (1) NO+O 3 NO 2 +O 2 (2) The monatomic oxygen present in (1) results from the dissociation of an ozone molecule due to interaction with ultraviolet radiation, as ozone absorbs radiation strongly in the ultraviolet range of wavelengths (240 to 320 nm). This monatomic oxygen would then recombine with diatomic oxygen to form additional ozone by: O+O 2 +M O 3 +M (3) where M is any third body in the chemical interaction. However, as can be seen from Equation (1), the presence of NO 2 converts the monatomic oxygen to diatomic oxygen and thus reduces the generation of ozone by Equation (3). In addition, the presence of NO converts ozone to diatomic oxygen by Equation (2). Thus, by Equations (1) and (2),

20 NO x reduces the amount of ozone in the atmosphere and in so doing reduces the benefits of ultraviolet light absorption by ozone molecules. Seinfeld [1] discusses how NO x is principally generated by combustion processes (such as those in gas turbines), and hence it is important to understand the fundamental aspects of NO x formation in gas turbine combustion processes and what means may be used to minimize NO x formation. Background The formation of NO x in combustion reactions has been studied extensively, particularly with applications towards flames used in gas turbines. There are two mechanisms which lead to the formation of nitric oxide, the first of which is the thermal mechanism. The thermal mechanism of NO formation was first proposed by Zeldovich, and consists of the following reactions: O+N 2 NO+N (4) N+O 2 NO+O (5) With the formation of NO, NO 2 may be formed by the following reaction: NO+HO 2 NO 2 +OH (formation) (6) NO 2 +O NO+O 2 (destruction) (7) Hori et al. [2] studied the process of NO conversion to NO 2 by Equation (6) in detail. They examined the formation of NO 2 from NO by mixing hot combustion gases with cool air with low levels of various hydrocarbon fuels. The combustion gases were at approximately 1400 K while the cool air was at ambient conditions. Their results showed that even small quantities of fuel in the exhaust gases can lead to a large portion of the 19

21 NO x being NO 2 by Equation (6). For C 3 H 8 at 40 ppm in the cool air being added to the exhaust gases, the percentage of NO 2 in the NO x increased from 24 percent to 90 percent. For methane, they examined equivalence ratios ranging from 0.76 to 1.2, with the maximum levels of NO 2 and NO x occurring at φ 1.0. The amount of fuel that was required to be added in order to bring the percentage of NO 2 to nearly 100 percent of the NO x varied with equivalence ratio, with the maximum being at φ 1.0. In all cases, an increase of fuel in the cool air lead to a large increase in the NO 2 found in the NO x. It should be noted that the levels of NO x did not vary with the amount of fuel added. Chemical kinetic calculations done by Hori et al. [2] showed that reaction (6) was the primary mechanism for the conversion of NO into NO 2, and that this reaction was promoted strongly by the use of fuels that readily decompose to form species which can lead to formation of HO 2. Their chemical kinetic calculations showed that, for a temperature of 1000 K and an initial concentration of 10 ppm of NO, hydrocarbon fuels such as C 3 H 8, C 2 H 4, and C 2 H 6 would promote the formation of HO 2 and thus NO 2 in the exhaust gases the most. CH 4 did not promote the formation of NO 2 as significantly. One important result of their work was to determine that the conversion of NO to NO 2 was heavily dependent on the temperature of the reaction, and that the temperature at which NO 2 formed most quickly varied with each hydrocarbon fuel. In their experiment, though the exhaust gases were at approximately 1400 K, NO conversion to NO 2 was found to occur at temperatures as low as 650 K with hydrocarbons present. The activation energy of Equation (4) is 315 kj/mole, which is considerably greater than the activation energy of many combustion reactions involving hydrocarbon fuels (approximately 160 kj/mole) [3,4]. The activation energy of Equation (5) is 20

22 considerably less, equal to 26 kj/mole [4]. This difference of activation energies means that the reaction rate of Equation (4) is much slower than that of combustion reactions. Reactions (4) and (5) are governed by an equation of the following form: k exp (-E a /RT) (8) Where k is the reaction rate constant, E a is the activation energy of the reaction, R is the gas constant, and T is the temperature. This mechanism is strongly dependent on the flame temperature, as can be seen by the temperature quantity in the exponent, and hence is referred to as the thermal mechanism. Glassman [4] discussed the fact that, due to the activation energy of reaction (4) being much higher than that of combustion reactions, the formation of NO by the thermal mechanism is heavily dependent on temperature. As previously mentioned, the formation of NO leads to the formation of NO 2 by Equation (3). The fact that NO x formation is related to flame temperature suggests that in lean flames (which are commonly used in modern gas turbines), the areas of greatest NO x formation will be the regions that have equivalence ratios larger than the average equivalence ratio for the entire flame. These regions, which are still fuel lean, will have a larger temperature due to the larger equivalence ratio, which in turn will lead to greater NO x formation. The second method of NO x formation is the prompt NO x mechanism. This mechanism was proposed by Fenimore [5] and consists of the following process: CH+N 2 HCN+N (9) C 2 +N 2 2CN (10) The N atoms formed in (9) could then form NO from reaction (5), and NO 2 could form from reaction (6). Measurements have been carried out on flat flame burners, and have 21

23 been observed most commonly in fuel rich hydrocarbon flames. Results have shown that the concentration of prompt NO x in methane flames can increase from nearly zero ppm in fuel lean flames (φ 0.7) to approximately 40 ppm in fuel rich flames (φ 1.3). Data collected has shown that the activation energy required for the prompt NO x mechanism to be on the order of 50 to 60 kj/mole. This lower activation energy leads to the prompt NO x mechanism having a much faster reaction rate (hence it is called prompt NO x ) than the thermal mechanism [4]. Fenimore speculated that reactions (9) and (10) could occur in flames which posses hydrocarbon fuels, as can be seen by the CH term in Equation (9). The fact that the concentration of hydrocarbon fuels contributes to this reaction suggests that the prompt NO x mechanism will indeed be more prevalent in fuel rich flames [5]. For fuel rich flames, the equivalence ratio is greater than one, and thus the temperature will be lower than if the fuel and air were mixed in stoichiometric proportions. This lower flame temperature means that formation of NO x by the thermal mechanism can be expected to be negligible compared to the NO x formed by the prompt mechanism. However, since gas turbine flames are usually fuel lean, the prompt mechanism is not likely to be significant compared to the thermal mechanism. As discussed in Glassman [4], the thermal mechanism can be controlled through reduction of flame temperature and residence time. Reducing regions that have equivalence ratios larger than the average value is a means of eliminating hot spots, thus reducing NO x formation in gas turbine combustors. In addition, a velocity field which limits the residence time of the reacting species in the reaction zone will also promote lower concentrations of NO x. Regions that would lead to high NO x formation due to larger equivalence ratios can be reduced through complete mixing of the fuel and air, and the velocity of the flow 22

24 field has important implications to the mixing of the reacting species. The free shear between regions with different mean velocities can create large values of RMS velocity, which will promote mixing of the fuel and air. In addition to influencing the RMS velocity, the mean velocity of the flow field determines the residence time of the reacting species, which is an important consideration since various components of the thermal and prompt mechanisms have different reaction rates. Thus, both the velocity and concentration fields of a flow are important with regards to the formation of NO x. Previous Work A significant amount of work has been done by a number of researchers into the formation of NO x under various conditions, and how NO x formation might be reduced. Steele et al. [6] studied the effect of various parameters (flame temperature, pressure, residence time, inlet temperature) on the formation of NO x under lean premixed combustion conditions. This work was done in two different atmospheric jet-stirred reactors with air as the oxidizer burning with various fuels in an effort to examine the fundamental aspects of NO x formation and the effects of various parameters such as temperature, pressure, and fuel species. For both reactors, the air and fuel were premixed in lean proportions prior to entering the reaction zone so that no mixing took place in the reaction zone. One of the reactors had a single center premixed fuel/air jet, while the second had eight diverging premixed fuel/air jets. The reason for the two configurations was to study the effects of inlet configuration on NO x formation. Due to the fact that the mixture was fuel lean, there was excess oxygen present in the exhaust gases. However, the molar concentration of oxygen will vary from one set of results to the next. Hence, 23

25 the results were corrected to a 15% O 2 dry basis. The results being on a dry basis simply means that the water formed by the combustion reaction was removed, as described in Turns [3]. Correcting to a 15% O 2 basis means that the mole fraction of the NO x in the exhaust gases was adjusted to correspond to what it would be if the molar concentration of oxygen in the exhaust gases were 15%. Correcting the results to 15% O 2 allows for easy comparison between the results of different experiments, which may have different mixtures and dilutions. The equation for normalizing the NO x concentration to a 15% O 2 basis is given by Equation (11): χ 15% = χ N mix (11) N mix, 15% where χ is the mole fraction of NO x measured in the exhaust gases, N mix is the total number of moles of all chemical species in the measured sample, N mix, 15% is the total number of moles of all chemical species at 15% O 2, and χ 15% is the calculated mole fraction of the NO x at 15% O 2. The data collected from the two reactors in the work of Steele et al. [6] was for a mean equivalence ratio ranging from 0.56 to Their results showed that there was no change in the ppm of NO x between the two inlet arrangements. However, other parameters studied did have an effect on NO x formation, such as the flame temperature, which depends on the equivalence ratio φ. When Steele et al. [6] examined the effects of temperature on NO x formation, the residence time of the reacting species kept at 3.5 ms under lean premixed conditions. They examined NO x levels generated using methane, propane, ethylene, and hydrogen/carbon monoxide as the fuel. Their results were taken over a temperature range 24

26 from approximately 1500 K to 1800 K. The levels of NO x formed at 1500 K were approximately 1.5 ppm for all four fuels. As the temperature increased to 1800 K, the NO x levels generated by the methane flame increased to 4 ppm while the NO x levels from the propane flame increased to 7 ppm. The NO x levels from the hydrogen/carbon monoxide and ethylene flames were approximately 5 and 6 ppm, respectively. The work of Steele et al. [6] showed that the difference in NO x levels between the various fuels increased with increasing flame temperature, particularly above 1700 K. From the raw data given in the appendix of their paper, NO x concentrations can be seen to increase in an exponential manner with temperature, which is consistent with the dependence of the thermal mechanism on temperature. However, the effects of unmixedness between the fuel and air were not analyzed in this experiment. As previously discussed, the mixing of fuel and air can have a significant impact on NO x formation. Since the flames investigated by Steele et al. were fuel lean, the prompt mechanism is not expected to have been significant in the NO x generated. Barnes and Mellor [7] used a numerical technique called a characteristic time model (CTM) to compute the sensitivity of NO x emissions to fuel/air unmixedness in the reaction zone. They used an unmixedness parameter that is calculated by dividing the standard deviation of the equivalence ratio by the time averaged equivalence ratio as given by Equation (12): s = σ φ /φ ave (12) where is s is the unmixedness parameter, σ φ is the standard deviation of the equivalence ratio, and φ ave is the mean equivalence ratio. The type of flame that was modeled was a lean premixed flame emanating from a combustor which had swirl vanes upstream of the 25

27 flame zone and a pilot nozzle along its centerline. As a result of the swirl vanes, there was a recirculation zone present in the flow. The model that was developed assumed that the main fuel/air mixture burned at an equivalence ratio less than one, and that this region of lean combustion surrounded the recirculation zone. Inside the recirculation zone, the fuel coming from the pilot nozzle burned with air near an equivalence ratio of one. One of the most important conclusions from this work is that turbulent eddies with equivalence ratios greater than the time averaged value contribute the most towards high concentrations of NO x. With a time averaged value of φ = 0.6, their numerical results predicted that approximately seventy percent of the NO x formation will be from eddies with φ > 0.6. In this work, NO x formation was quantified by the use of an emission index (grams of NO x formed per kilogram of fuel burned). Their results also showed that for an unmixedness parameter of s = 0.15, NO x production is 15% greater than for the same fuel flow rate under conditions in which the fuel and air are perfectly mixed. These numerical results further validate the importance of the thermal theory in that the greatest NO x formation is seen to come from relatively fuel rich eddies. Barnes and Mellor [8] compared the predictions of their CTM to experimental data in an effort to quantify the effects of unmixedness in operational gas turbines and estimate the unmixedness of the fuel and air in the reaction zone, where unmixedness for their research was defined by Equation (12). The data were collected from the exhaust gases at the exhaust plane of a lean premixed gas turbine combustor under fired conditions, and consisted of the NO x and CO emissions. CO emissions were measured so that the effect of unmixedness on them as well as on NO x emissions could be analyzed as CO is also a pollutant of concern. The exhaust gases were sampled simultaneously over 26

28 the entire exhaust plane and analyzed. The NO x data gathered represented the area average value of NO x for the exhaust plane. Because of this area averaging of the exhaust gases, it was not possible to examine the fluctuations of NO x concentration at different points across the exhaust plane, or to examine the temporal variations of NO x at any point in the exhaust plane. The equivalence ratio was varied from 0.5 to 0.8. They compared their test data to the CTM for various values of unmixedness in the reaction zone. They plotted their test data along with CTM data, and found that the test data plot collapsed onto the CTM data plot for an unmixedness value of s = 0.35 very well. However, because they were unable to gather fuel concentration data from the reaction zone, they cannot confirm this result. The fact that the measurements were taken from the exhaust gases also may lead to inaccuracy since the levels of NO x and CO in the exhaust gases may be different than those in the reaction zone due to changing chemical equilibrium and the oxidation of CO. As previously mentioned, it was desired to determine the effects of unmixedness on CO as well as NO x. Fric [9] also took measurements of the unmixedness of a fuel/air mixture and correlated its effect on NO x emissions, although the unmixedness parameter used by Fric was different than that used by Barnes and Mellor [7,8]. The unmixedness parameter used by Fric was calculated as follows: U = c 2 (13) (c ave (1-c ave )) where U is the unmixedness parameter, c 2 is the variance of the fuel concentration fluctuations and c ave is the time averaged fuel concentration. A coflow jet arrangement was used to study four different cases, varying from purely premixed to purely unmixed flow, with a mean equivalence ratio of 0.5 in all cases and at atmospheric pressure. It 27

29 was shown that for levels of unmixedness on the order of 10%, NO x production was approximately double that of the case in which there is no unmixedness. One aspect of fuel/air mixing that was not addressed by Fric [9] was mixing in regions other than the reaction zone, and the potential importance of this mixing is a significant motivation for the work proposed here. Mongia et al. [10] used an unmixedness parameter defined as in Equation 10 in the same way as Fric [9] to quantify the effect of unmixedness at different equivalence ratios and different pressures in a lean premixed burner in which methane mixed with air. Their results qualitatively confirmed the results of Fric [9] with regards to NO x production versus unmixedness values, and also showed that for low levels of unmixedness (U.0002), the production of NO x is independent of pressure. However, they found that for higher levels of unmixedness, NO x production increases with pressure. Their data showed that for typical gas turbine conditions (T= 1800 K, φ = 0.5), the NO x produced at 20 atm is approximately three times larger than for 1 atm with an unmixedness value of U = at both 1 atm and 20 atm. This experiment did not incorporate the swirled flow that many modern gas turbines do and the examination of the development of the mixing profile of a swirling flow is important to understanding the formation of NO x in gas turbines. Frey et al. [11] experimentally investigated the effect of the length of the premixed and turbulence on the mixing of fuel into air in a gas turbine premixer. The premixer they examined had a length of 8.89 cm and had 16 spray bars located downstream of the swirl vanes, which had a swirl angle of 48 o. The spray bars consisted of cylindrical bars with multiple holes in them through which the fuel flowed into the 28

30 swirling air. The fuel was simulated using 1 µm size aluminum oxide particles. The concentration of the simulated fuel was measured using a Mie scattering imaging technique, which used a pulsed laser to illuminate the seeder particles in the air in the premixer. Based on the intensity of the scattered light, the concentration of the particles can be calculated. This allows for examination of the concentration of fuel across an entire plane of the airflow. Frey et al. [11] used this Mie scattering imaging technique to examine the concentration field at the exit of the premixer. The turbulence, which was measured by hot wire anemometry, was created by the use of an inlet grid. Their experimental setup used two different inlet grids, which generated turbulence levels of 2.5% and 8%, with the baseline case being no grid and therefore laminar flow at the entrance plane of the premixer. It was found that the increased turbulence from these grids reduced the unmixedness of the flow, but not significantly. For their baseline case (no inlet grids), the value of the unmixedness parameter (defined the same as Barnes and Mellor [7]) was s = With the inlet grids in place, this improved only to 0.27 for the 2.5 % turbulence level and to 0.23 for the 8% turbulence level. They speculated that the turbulence due to the wakes of the swirl vanes and the spray bars were sufficiently strong such that the turbulence due to the inlet grids was not significant to the mixing of the fuel and air. However, as the free stream turbulence did have an effect on the unmixedness of the flow (and therefore would have an effect on the NO x formed from combustion of such a flow), analysis of the turbulence of a flow field could yield insight into the mixing process of fuel and air in a gas turbine mixer. Frey et al. [11] did not present any power spectrum data of the turbulence they measured, and analyzing the power spectrum of the 29

31 turbulence of a mixing flow will help to illustrate the contribution of free stream turbulence to the mixing of fuel and air. More significant reduction in unmixedness observed by Frey et al. [11] was due to increased premixer length. The two different extensions of 2.5 cm and 5 cm increased the length of the mixing channel by 57% and 116%, respectively. The 2.5 cm extension reduced the unmixedness from s = 0.29 to 0.23, and the 5 cm extension reduced the unmixedness to The increased residence time of the fuel and air in the premixer obviously will promote further mixing, and the work done by Frey et al. [11] highlights the importance of this parameter. Their work, however, did not examine the velocity field of the test hardware. The velocity field has great importance to the mixing of the fuel and air, and this was discussed in their conclusions. Due to the fact that they used a Mie-scattering imaging technique to measure the concentration of the simulated fuel, they were only able to analyze the concentration profile at the exit of the premixer, and were not able to analyze the development of the mixing profile throughout the length of the premixer. Analysis of the development of the mixture profile along with the velocity profile could yield a greater understanding of the dependence of fuel concentration on various parameters. Stufflebeam et al. [12] used an acetone fluorescence method to examine the fuel air ratio across the flow fields of premixed nozzles. The concentration of the acetone could be calculated by measuring the intensity of the scattered light, which is similar to the work done by Frey et al. [11]. The acetone was injected both through a pilot nozzle located along the centerline of the flow (this gave a annular cross section to the flow), and through two main lines which supply fuel to the airflow parallel to the direction of 30

32 the airflow, and without the use of swirl vanes. The air was seeded with 2% acetone by weight, and the fuel/air momentum flux ratio at the location of fuel (air seeded with acetone) was kept identical to that in an operational gas turbine. For this work, unmixedness was defined as the standard deviation of the fuel/air ratio divided by the average fuel to air ratio, which is the definition given by Equation (12). They showed that, among other parameters the fuel/air ratio was strongly dependent on the geometry of the nozzle. It was illustrated that by varying parameters such as the number of holes through which the fuel is injected and the momentum ratio of the fuel to the air, significant reductions in unmixedness can be realized. Their results showed that by injecting fuel at the upstream end of the pilot nozzle, the fuel/air ratio was largest near the inner wall of the annular geometry. Conversely, injecting the fuel near the downstream end of the pilot nozzle resulted in the fuel/air ratio being largest near the outer wall. These results were taken at the exit of the nozzles. The effect of fuel injection geometry on the development of the concentration profile was not examined. Acetone laser induced fluorescence (LIF) was also used by Thomsen et al. [13] to determine the effect of the orientation angle of the fuel jet to the flow on mixing between the fuel and air in a lean premixed gas turbine premixer that incorporated swirl vanes. The LIF images taken showed the mixing of the fuel and air along the premixer length, which was 3.5 cm. The fuel was injected through 16 fuel delivery tubes that were located in between the swirl vanes. The swirl vanes themselves were located at the entrance of the swirler. The angle of the fuel jet to the flow was varied from 60 o to 60 o in increments of 20 o. In this experiment, a negative angle means that the fuel is injected against the flow of the air. Analysis of their results showed that an angle of -20 o yielded 31

33 the most uniform fuel/air mixture (i.e., the smallest levels of unmixedness). Their results also showed that for the case of a 0 o fuel injection angle, the mixing of fuel and air was nearly complete prior to the flow entering the reaction zone. This work highlights the importance of fuel injection angle to mixing of fuel and air, and the LIF images taken illustrate the importance of premixer length in swirled flows. Work has also been done to numerically analyze the mixing of fuel and air in turbulent non-premixed flows. Komori et al. [14] used an unmixedness parameter similar to that used by Barnes and Mellor [6], which they refer to as a segregation parameter. It is defined by Equation (14): α = c c A A c B c B (14) where c A is the mean concentration of species A, c B is the mean concentration of species B, and c is the cross correlations between the fluctuation of the concentration c A B of species A and B, and α is the segregation parameter. Their findings showed that increasing shearing in the flow (such as the free shear caused by a recirculation zone) can significantly improve mixing as evidenced by smaller values of the segregation parameter. Improved mixing will improve combustion by reducing fuel unmixedness and therefore NO x production by the thermal mechanism. Mori et al. [15] showed that for the lean premixed gas turbine combustor that they examined, a second order Reynolds Stress Model was successful in modeling the concentration of methane across the profile of the flow field. The hardware that they attempted to model was a lean premixed gas turbine combustor incorporating an axial swirler and a pilot nozzle along the centerline of the combustor. The air entered the 32

34 combustor at one atmosphere and at ambient temperature by two different paths. Some of the air was injected into the flow along the axis of the combustor by the pilot nozzle (there were swirl vanes inside the pilot nozzle to swirl the air passing through it), while premixed gas pipes located upstream of the swirl vanes injected the fuel/air mixture into the main airflow. The air passing through the gas pipes and swirl vanes enters the combustor at a diagonal angle to the centerline of the combustor. Instead of fuel being used to seed the airflow for concentration measurements, small oil particles were used and their concentration measured by a planar Mie scattering technique, as was done by Frey et al. [11]. The results presented by Mori et al. [15], which consisted of simulated fuel concentration data taken at the exit of the combustor, showed that the second order Reynolds Stress Model correctly predicted the radial profile of mass fraction of fuel to within five to ten percent except at the outer ten percent of the radius. As with other concentration work done, such as by Stufflebeam [12] and Thomsen [13], the data was gathered only at the exit of the test hardware. No data was gathered throughout the mixing region of the hardware, and knowledge of the mixing behavior in this region could lead to insights into the design of better gas turbine mixers. Some work has been done with regards to the velocity field of a reacting flow by Puri et al. [16]. They used a laser Doppler velocimeter to analyze the axial and tangential velocity components from a lean premixed combustor for an industrial gas turbine. The purpose of this work was to confirm their numerical modeling of the flow field, in which they modeled the velocity fields and the mixing of fuel and air as the flow progressed through a gas turbine combustor. The lean premixed combustor that they examined incorporated a radial swirler upstream of a converging-diverging nozzle section. The 33

35 reaction zone was located in the diverging section of the nozzle. The combustor examined could be configured with several different fuel hole patterns and swirler configurations, and atmospheric tests were performed with a concentration probe to determine which arrangement yielded the most complete mixing of fuel and air. The results of combustion tests showed that the configuration that gave the most uniform fuel concentration at the exit of the combustor also had the lowest concentration of NO x, with a NO x concentration of 4 to 6 ppm at φ = Under operational conditions, the amount of NO x generated was shown to vary between the different configurations (and therefore the different concentration profiles). With less uniform mixing, the levels of NO x generated were approximately 10% higher. However, the amount of NO x formed was shown to depend more strongly on the flame temperature. For all configurations, the NO x concentration was approximately 6 ppm at 1435 K, while at 1655 K the NO x concentration was approximately 40 ppm. The combustion temperature was varied by changing the temperature of the air entering the combustion section. These results show the importance of both flame temperature and fuel/air mixing in NO x formation. Due to the swirling nature of the flow and the flow expansion in the combustor examined by Puri et al. [16], there was a recirculation zone present in the flow field. While the width and length of the recirculation zone was described and its importance to the stability of the flame in their radial swirler was discussed, no detailed information was given as to how the velocity field corresponded to or affected the concentration field of the flow. Numerical modeling had shown that the wakes of the fuel injection tubes resulted in rapid mixing of the fuel and air, but the effect these wakes was not discussed extensively. Axial velocity profiles at various locations throughout the combustor were 34

36 given, but none for the tangential velocity component. In addition, no RMS velocity data was given for either the axial or tangential velocity components. Since the flow in gas turbine combustors is a highly turbulent region due to the free shear between velocity layers, the RMS velocity will play an important part in the mixing between fuel and air. The importance of the RMS velocities to mixing in the free shear layers due to the presence of a recirculation zone was shown by Ahmed et al. [17]. They used a two component (axial and tangential velocity component) argon-ion LDV to measure airflow emanating from an axial dump combustor which had 20 swirl vanes at an angle of 45 o and a fuel injector along its centerline. The airflow was seeded with titanium dioxide particles so that the mean velocities as well as the turbulence intensity could be measured. They found that large velocity gradients and turbulence levels are important to the improvement of the mixing of the species, particularly at axial locations much less than one combustor radius downstream. Due to the fact that the flow emanating from the combustor expanded, there was a significant radial component near the exit of the combustor. It was found that the high turbulence of the flow, combined with the impinging of the airflow on the wall of the flow channel, lead to the radial component diminishing at locations less than one combustor radius downstream of the combustor exit. After the dissipation of the radial velocity component, the flow was twodimensional in nature. Ahmed et al. [17] concluded that confined swirling flows were highly dissipative, particularly in the region near the swirl vanes and that rapid mixing of the flow would be an important consequence of this dissipative behavior. Much of the work done in regions near swirl vanes or turbine blades has focused on heat transfer phenomenon rather than fuel/air mixing. Camci and Arts [18] studied the 35

37 effect of incidence on the flow around a set of swirl vanes, in addition to studying the effect of mass ejection from the blades on the secondary flow around the vanes. The focus of this work, and many other papers like it, was to examine the effect of various parameters on the heat transfer behavior of the vanes. The mass ejection from the vanes that was studied was for the purposes of film cooling, rather than for providing fuel for a combustion reaction. In particular, the changing location of the stagnation point of the flow on the swirl vanes was analyzed. The motivation for studying the effect of incidence on the flow behavior was the changing incidence of vanes in operational turbines due to the wakes of vanes at upstream locations. It was found that the ejection of mass from the vane affected the flow transition and separation within distances of 20% of the chord length downstream of the mass ejection location. The effect of mass ejection from vanes is an important consideration in any work which involves mass ejection from vanes. While the work done by Camci and Arts [18] illustrated the impact of mass ejection from the vanes on flow turbulence, their work focused on heat transfer and the flow behavior downstream of the swirl vanes was not analyzed. Friedrichs et al. [19] also showed that the mass ejection from the vane can significantly alter the boundary layer flow. This work also focused largely on heat transfer effects, and the mass being ejected was ejected at an angle of 30 o to the surface of the blade. They found that the mass ejection can significantly affect the losses of the flow through the vanes. Depending on the inlet Mach number, the flow losses due to mass ejection varied from 6% to 8% of the total losses. The separation behavior of the flow was also influenced. Mass ejection upstream of the point of flow separation was shown to delay the separation, while mass ejection downstream of the separation point 36

38 did not alter the flow behavior significantly. Thus, from the work of Friedrichs et al. [19] and Camci and Arts [18], it can be seen that mass ejection from the vanes can significantly affect the flow around the vanes. Hence, mass ejection for the purposes of injecting fuel into the air flow through a gas turbine mixer could also be expected to influence the flow around the swirl vanes, and hence the mixing of the fuel and air. Such mass ejection would have an impact on the turbulence of flow around such vanes, and since this is the case the spectrum of the turbulence will be important to any effort to understand the behavior of flow emanating from swirl vanes. The importance of the spectrum of turbulence is discussed by Tennekes and Lumley [20]. They discussed how energy is typically imparted to turbulence by large scale structures, such as swirl vanes. However, the dissipation of this energy typically occurs at relatively small scales. From the above literature discussed, it is clear that the mixing of fuel and air has a significant effect on the emissions from lean swirled flames emanating from gas turbine mixers. Thus, it is desirable to better understand the mixing process in swirling flow through a gas turbine mixer. In order to do this, the development of the velocity and fuel concentration profiles in two gas turbine mixers was analyzed. The two gas turbine mixers had different swirl angles, and the primary purpose of this work was to develop a greater understanding of the effect of swirl vane angle on the velocity field, the mixing of fuel and air, and their implications towards NO x formation. 37

39 EXPERIMENTAL METHODS AND SETUP It has clearly been shown, by Fric [9] and also Barnes and Mellor [7,8] and others, that both the spatial and temporal unmixedness between the fuel and air has strong implications towards NO x emissions from lean premixed flames. It has also been shown that the velocity field (both the mean and RMS velocities) has significant effect on the mixing of fuel and air, and that the length of the mixing channel and hardware geometry also can affect the development of the concentration profile. This makes it important to know the concentration and velocity fields of a fuel/air flow in order to better understand the mixing of the fuel and air. Therefore, a concentration probe was used to examine the mixing of methane into air from two different lean premixed gas turbine mixers in conjunction with the use of a Laser Doppler Velocimeter (LDV) to measure the velocity field (note that the LDV and concentration probe were not used simultaneously). From the concentration probe, both the time averaged and temporal variation in concentration (for the purposes of an FFT analysis) was measured. From the LDV, axial and tangential velocity as well as axial and tangential RMS velocity data was measured. The only significant geometric difference between the gas turbine mixers was the angle of the swirl vanes (45 o versus 55 o ). The swirler nozzle with a 45 o angle will be referred to as Module 3, and the swirler nozzle with a 55 o angle will be referred to as Module 4, following their nomenclature from Siemens-Westinghouse Power Corporation, which is the designer of the two swirler nozzles examined. The use of these two modules allowed for an examination of the effect of the swirl angle on both the velocity field and concentration field. Table 1 below 38

40 shows the relevant geometric parameters of the two gas turbine mixers examined in this work Table 1. Geometric parameters of swirler nozzles. Parameter Module 3 Module 4 Outer radius 47 mm 49.5 mm Inner radius 23.5 mm 23.5 mm Swirl Angle 45 o 55 o Fuel Injection Profile Flat Flat The concentration and velocity measurements were taken downstream of the swirl vanes throughout the region in which the concentration and velocity profiles of the flow field are changing. The configuration of the test hardware, as well as its relevant dimensions, can be seen in Figure 2 below. The region in which both velocity and concentration measurements will is referred to as the aft shroud. The aft shroud of Module 3 has a hydraulic diameter of 47 mm, and the range over which measurements were taken was 0.89 to 3.02 hydraulic diameters downstream of the trailing edge of the swirl vanes. For the aft shroud of Module 4, the hydraulic diameter is 52 mm and the range over which measurements were taken was 0.87 to hydraulic diameters downstream of the trailing edge of the swirl vanes. The distance between the trailing edge of the swirl vanes and the data collection locations will be labeled as. 39

41 A diagram of a swirl vane and its associated geometric parameters can be seen in Figure 1. The geometric parameters described in Figure 1 are the chord length of the swirl vane (c v ), the height from the leading edge to the trailing edge (h v ), the maximum thickness (t max ), and the length along the camber line of the swirl vane (l camber ). The value of each of these parameters is listed in Table 2 for both Module 3 and 4. t max l camber h v c v Figure 1. Diagram of swirl vane geometric parameters. Table 2. Geometric parameters of swirl vanes. Parameter Module 3 Module 4 c v 25.4 mm 22.2 mm h v 15.9 mm 20.6 mm t max 6.4 mm 6.4 mm l camber 28.6 mm 28.6 mm It was intended that = 2.40 would be one of the data collection locations for Module 4, however there was a change in outer diameter in the acrylic aft shroud of 40

42 Module 4 at this location. This change in outer diameter resulted in the LDV not being able to receive Doppler signals due to the changing optical properties that accompanied the change in outer diameter at = In order to be able to examine the development of the flow field, it was decided to collect data at = 2.21 and = in Module 4. These locations differed by only 10 mm from the two most downstream data locations in Module 3, and are the reason that the range of hydraulic diameters over which data was collected differs for Modules 3 and 4. At the far end of the acrylic tube shown in Figure 2 is a fan motor which draws the airflow through the test rig. The bolt on feed flange shown in Figure 2 is the path through which methane was supplied to the test hardware for concentration measurements. Since it was desired to keep methane concentrations very low to prevent a risk of combustion, the methane was diluted with air, and the momentum flux ratio between the air flowing through the swirler and the air/methane mixture being injected was set to match operational conditions. The methane/air mixture is injected into the airflow through holes in the lower (pressure) side of the swirl vanes As previously mentioned, the LDV and the concentration probe were used separately to collect data in this work, and therefore the discussion of their use will be done separately as well. 41

43 Figure 2. Test rig with swirler mounted into acrylic tube.

44 Laser Doppler Velocimeter In order to allow the lasers of the LDV to pass through the wall of the aft shroud into the airflow, a groove was machined into the aft shroud and a piece of glass inserted into the groove. The glass was covered with an antireflective coating, the purpose of which is to prevent laser light reflections off of the glass from interfering with reception of Doppler signals by the receiving optics. A diagram of the lasers from the LDV entering the aft shroud through the antireflective glass is shown in Figure 3, and Figure 4 shows a diagram of the LDV in relation to the test rig. A picture of the Module 3 mounted into the test rig is shown in Figure 5. The machining of a groove was not necessary for the case of Module 4, as the aft shroud that was designed for it was made of acrylic, and therefore was optically accessible by the LDV. The LDV was configured to collect Doppler signals in backscatter mode. This was done because the presence of the swirler hub along the centerline of the hardware prevented light signals from being seen by the receiving optics had they been placed on the opposite side of the aft shroud. As can be seen in Figure 4, the receiving optics are mounted onto the LDV and receive Doppler signals generated in the probe volume of the LDV. The directions that the Doppler signals travel in is opposite to the direction that the lasers travel in, hence the term backscatter.

45 Aft Shroud Module #3 Swirler Hub Probe Volume (l =.655 mm, d =.131 mm) Antireflective Glass Window (t = 3.3 mm, h = 13 mm) Laser beams Focusing Lens Figure 3. Test hardware with lasers passing through antireflective glass.

46 Figure 4. Diagram of LDV and test rig. 45

47 Aft Shroud 45 o arcs Swirl Vanes Groove with antireflective glass inserted Figure 5. Module 3 mounted in test rig. In Figure 5 (a photograph of Module 3), it can be seen that arcs were machined into the leading edge flange of the aft shroud. The purpose for doing this was to allow for the aft shroud to be rotated around its axis and secured at a given angle. This allowed for examination of the axial and tangential profiles at various angles of rotation. By rotating the hardware in the θ direction, the wakes of the swirl vanes can be analyzed at different angular locations. Each arc machined into the aft shroud encompasses an angle of 45 o. The mounting plate to which the test hardware is bolted has also has 45 o arcs machined into it. Thus, the test hardware can be rotated to the desired angle and the aft shroud rotated in the opposite direction so as to always keep the antireflective glass window facing the LDV. This was only a problem with Module 3, as it was not completely optically clear, whereas the aft shroud of Module 4 was. A picture of Module 4 mounted in the test rig can be seen below in Figure 6. Data was taken for both modules at the same θ angles. 46

48 45 o arcs Swirl Vanes Optically Clear Aft Shroud Figure 6. Module 4, with optically clear aft shroud, mounted in test rig. In using the concentration probe and the LDV, it is important to consider the spatial and temporal resolution of the data of these two systems. The LDV uses a 100 mm beam spacer with a 250 mm focal length. This gives the following dimensions for the probe volume, shown below in Table 3: Table 3. LDV parameters (see Figure 7). Parameter l m d m d f λ Value mm mm 1.38µm nm 47